Military applications are a small subset and are unaffected by copyright issues. Applications can be trained in secrecy.
The copyright, plagiarism and unemployment issues are entirely disjoint from the national security issues. If North Korea trains a chat bot using material that is prohibited for training by a special license, so what? They already don't respect IP.
I can almost 100% guarantee, with regulation, you'll see all the same loss of jobs and whatnot, but only the chosen few who are licensed will hold the technology. I'm old enough to have seen the interplay of corporations, the government, and regulatory capture, and see what that's done to the pocketbook of the middle class.
No. Thank. You.
I don't trust these corporate hucksters one bit. As I've said in a previous comment: if they want to demonstrate their earnest benevolence, why don't they work on regulation to reign in the previous humanity-enslaving mess they created - commercial mass surveillance.
There are literally billions of people that can be empowered by these tools. Imagine what will result when the tens of thousands of “one in a million” intellects are given access to knowledge that only the richest people have had. Rich incumbents have reason to be worried.
The dangers of tools like these are overblown. It was already possible for smart actors to inflict massive damage (mass poisoning, infrastructure attacks, etc). There are so many ways for a person to cause damage, and you know what? Few people do it. Most Humans stay in their lane and instead choose to create things.
The real thing people in power are worried about is competition. They want their monopoly on power.
I’m really optimistic of legislation like Japan’s that allows training of LLM on copyrighted material. Looking for great things from them. I hope!
Sure, sometimes it leads me astray but generally it keeps course.
In theory with the internet I have access to most of the data humanity created. But that's not much different than throwing me on a raft in the middle of the ocean and going 'here, all the water you want.
Especially with search just getting so crappy and seo spam filled. It's something I can use to refine the knowledge I have by helping to focus the torrent.
Fill your boots while you can, everyone. The great consolidation is here!
- AI may be literally dangerous technology (i.e. Skynet)
- AI may cause mass unemployment
- AI may not be dangerous but it's a critical technology for national security (We can't afford an AI gap.)
- Generative AI may be violating copyright (which is really just a government policy question)
Military applications are a tiny subset of evil that can be done with intelligence, artificial or otherwise. So much of the global economy is based on IP, and AI appears to be good at appropriating it and shoveling out near infringements at a breathtaking scale. Ironically, AI can paraphrase a book about patent law in a few minutes... and never really understand a word it wrote. At the moment AI may be an extistential threat to IP based economies... which is certaintly as much of a national security threat as protecting the water supply.
> "If we don't do it, the evil people will do it anyway" is not a good argument.
This is a good argument if the cow was still in the barn. At this moment, we're all passengers trying to figure out where all of this is going. It's change, and it's easy to be afraid of it. I suspect, though, just like all changes in the past AI could just make life better. Maybe.
https://en.wikipedia.org/wiki/List_of_parties_to_internation...
But in general copyright doesn't apply to governments, even here in the US. The North Korean government can violate copyright all it wants to, its subject citizens can't, though. https://www.natlawreview.com/article/state-entity-shielded-l...
I think it would likely be a constant competition rather than take over. Why is it destructive?